Methods and systems for detecting frauds by utilizing spend patterns of payment instruments of user

Abstract

Embodiments provide methods and systems for detecting frauds in payment transactions made by payment instrument using spend patterns of multiple payment instruments associated with user. The method performed by server system includes receiving payment transaction data associated with first payment instrument including information of payment transaction performed at particular merchant. Method includes generating multivariate payment transaction sequence associated with one or more second payment instruments of user. Method includes predicting simulated univariate payment transaction sequence associated with the first payment instrument based on first neural network model and the multivariate payment transaction sequence. Method includes providing simulated univariate payment transaction sequence and real univariate payment transaction sequence of first instrument to second neural network model. Method includes determining that payment transaction is fraudulent based, at least in part, on comparison of simulated univariate payment transaction sequence and real payment transaction sequence by the second neural network model.

We claim: 
     
         1 . A computer-implemented method comprising:
 receiving, by a server system, payment transaction data associated with a first payment instrument of a user, the payment transaction data comprising information of a payment transaction performed at a particular merchant;   generating, by the server system, multivariate payment transaction sequence associated with one or more second payment instruments of the user, the multivariate payment transaction sequence representing past payment transactions accessed from a database, and the past payment transactions performed using the one or more second payment instruments at the particular merchant over a threshold period of time;   predicting, by the server system, a simulated univariate payment transaction sequence associated with the first payment instrument based, at least in part, on a first neural network model and the multivariate payment transaction sequence;   providing, by the server system, the simulated univariate payment transaction sequence and a real univariate payment transaction sequence generated based on the receipt of the payment transaction data associated with the first payment instrument of the user to a second neural network model; and   determining, by the server system, that the payment transaction is fraudulent based, at least in part, on a comparison of the simulated univariate payment transaction sequence and the real univariate payment transaction sequence by the second neural network model.   
     
     
         2 . The computer-implemented method as claimed in  claim 1 , wherein the first neural network model and the second neural network model are incorporated in a generative adversarial network (GAN) model and wherein the first neural network model is a generator neural network model and the second neural network model is a discriminator neural network model. 
     
     
         3 . The computer-implemented method as claimed in  claim 1 , further comprising:
 aggregating, by the server system, the payment transaction and previous spendings associated with the first payment instrument performed at the particular merchant over the threshold period of time, to generate the real univariate payment transaction sequence of the first payment instrument.   
     
     
         4 . The computer-implemented method as claimed in  claim 2 , wherein determining that the payment transaction is fraudulent comprises:
 identifying, by the discriminator neural network model, a deviation value between the simulated univariate payment transaction sequence and the real univariate payment transaction sequence; and   determining the payment transaction being fraudulent when the deviation value is greater than a predetermined threshold value.   
     
     
         5 . The computer-implemented method as claimed in  claim 2 , further comprising:
 aggregating, by the server system, one or more payment transactional features corresponding to the past payment transactions of the one or more second payment instruments to generate the multivariate payment transaction sequence.   
     
     
         6 . The computer-implemented method as claimed in  claim 5 , wherein the one or more payment transactional features comprise an amount of spend, a merchant category code (MCC), merchant risk profile, a frequency of purchase at the particular merchant, and an average amount of purchase at the particular merchant. 
     
     
         7 . The computer-implemented method as claimed in  claim 2 , wherein, in a training phase, the GAN model is trained by:
 accessing payment transaction data of a plurality of payment instruments associated with the user, the payment transaction data comprising information of past payment transactions performed at a plurality of merchants using the plurality of payment instruments within a particular time duration;   selecting a first payment instrument from the plurality of payment instruments to learn spending patterns of the first payment instrument at a specific merchant;   generating multivariate payment transaction sequence of one or more second payment instruments from the plurality of payment instruments by aggregating payment transactional features of the past payment transactions associated with the one or more second payment instruments on timely basis;   transmitting the multivariate payment transaction sequence and a merchant flag vector to the generator neural network model for generating a simulated univariate payment transaction sequence associated with the first payment instrument; and   transmitting the simulated univariate payment transaction sequence and a real univariate payment transaction sequence associated with the first payment instrument to the discriminator neural network model.   
     
     
         8 . The computer-implemented method as claimed in  claim 7 , wherein the GAN model is further trained by:
 determining, by the discriminator neural network model, a deviation value between the simulated univariate transaction sequence and the real payment transaction sequence associated with the first payment instrument; and   updating neural network weights of the generator neural network model based, at least in part, on the deviation value.   
     
     
         9 . The computer-implemented method as claimed in  claim 7 , wherein the merchant flag vector is utilized for conditioning input of the generator neural network model based on the specific merchant. 
     
     
         10 . A server system comprising at least one processor in communication with at least one memory, the at least one processor configured to:
 receive payment transaction data associated with a first payment instrument of a user, the payment transaction data comprising information of a payment transaction performed at a particular merchant;   generate multivariate payment transaction sequence associated with one or more second payment instruments of the user, the multivariate payment transaction sequence representing past payment transactions accessed from a database, and the past payment transactions performed using the one or more second payment instruments at the particular merchant over a threshold period of time;   predict a simulated univariate payment transaction sequence associated with the first payment instrument based, at least in part, on a first neural network model and the multivariate payment transaction sequence;   provide the simulated univariate payment transaction sequence and a real univariate payment transaction sequence generated based on the receipt of the payment transaction data associated with the first payment instrument of the user to a second neural network model; and   determine that the payment transaction is fraudulent based, at least in part, on a comparison of the simulated univariate payment transaction sequence and the real univariate payment transaction sequence by the second neural network model.   
     
     
         11 . The server system of  claim 10 , wherein the first neural network model and the second neural network model are incorporated in a generative adversarial network (GAN) model and wherein the first neural network model is a generator neural network model and the second neural network model is a discriminator neural network model. 
     
     
         12 . The server system of  claim 10 , wherein the at least one processor is further configured to:
 aggregate the payment transaction and previous spendings associated with the first payment instrument performed at the particular merchant over the threshold period of time, to generate the real univariate payment transaction sequence of the first payment instrument.   
     
     
         13 . The server system of  claim 11 , wherein to determine that the payment transaction is fraudulent, the at least one processor is further configured to:
 identify, by the discriminator neural network model, a deviation value between the simulated univariate payment transaction sequence and the real univariate payment transaction sequence; and   determine the payment transaction being fraudulent when the deviation value is greater than a predetermined threshold value.   
     
     
         14 . The server system of  claim 11 , wherein the at least one processor is further configured to:
 aggregate one or more payment transactional features corresponding to the past payment transactions of the one or more second payment instruments to generate the multivariate payment transaction sequence.   
     
     
         15 . The server system of  claim 14  wherein the one or more payment transactional features comprise an amount of spend, a merchant category code (MCC), merchant risk profile, a frequency of purchase at the particular merchant, and an average amount of purchase at the particular merchant. 
     
     
         16 . The server system of  claim 11  wherein, in a training phase, the GAN model is trained by the at least one processor further configured to:
 access payment transaction data of a plurality of payment instruments associated with the user, the payment transaction data comprising information of past payment transactions performed at a plurality of merchants using the plurality of payment instruments within a particular time duration; 
 select a first payment instrument from the plurality of payment instruments to learn spending patterns of the first payment instrument at a specific merchant; 
 generate multivariate payment transaction sequence of one or more second payment instruments from the plurality of payment instruments by aggregating payment transactional features of the past payment transactions associated with the one or more second payment instruments on timely basis; 
 transmit the multivariate payment transaction sequence and a merchant flag vector to the generator neural network model for generating a simulated univariate payment transaction sequence associated with the first payment instrument; and 
 transmit the simulated univariate payment transaction sequence and a real univariate payment transaction sequence associated with the first payment instrument to the discriminator neural network model. 
 
     
     
         17 . The server system of  claim 16  wherein the GAN model is further trained by the at least one processor further configured to:
 determine, by the discriminator neural network model, a deviation value between the simulated univariate transaction sequence and the real payment transaction sequence associated with the first payment instrument; and 
 update neural network weights of the generator neural network model based, at least in part, on the deviation value. 
 
     
     
         18 . The server system of  claim 16  wherein the merchant flag vector is utilized for conditioning input of the generator neural network model based on the specific merchant.